[PATCH v4] kernel/watch_queue: NULL the dangling *pipe, and use it for clear check

[PATCH v4] kernel/watch_queue: NULL the dangling *pipe, and use it for clear check

[Siddh Raman Pant]

NULL the dangling pipe reference while clearing watch_queue.

If not done, a reference to a freed pipe remains in the watch_queue,
as this function is called before freeing a pipe in free_pipe_info()
(see line 834 of fs/pipe.c).

The sole use of wqueue->defunct is for checking if the watch queue has
been cleared, but wqueue->pipe is also NULLed while clearing.

Thus, wqueue->defunct is superfluous, as wqueue->pipe can be checked
for NULL. Hence, the former can be removed.

Signed-off-by: Siddh Raman Pant <code@siddh.me>
---
Changes in v4:
- Drop preceeding kerneldoc-changes patch and change appropriately.

Changes in v3 (8 Jan 2023):
- Minor rephrase of comment before NULLing in watch_queue_clear().

Changes in v2 (6 Aug 2022):
- Merged the NULLing and removing defunct patches.
- Removed READ_ONCE barrier in lock_wqueue().
- Better commit messages.

 include/linux/watch_queue.h |  3 +--
 kernel/watch_queue.c        | 12 ++++++------
 2 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/include/linux/watch_queue.h b/include/linux/watch_queue.h
index fc6bba20273b..45cd42f55d49 100644
--- a/include/linux/watch_queue.h
+++ b/include/linux/watch_queue.h
@@ -38,7 +38,7 @@ struct watch_filter {
 struct watch_queue {
 	struct rcu_head		rcu;
 	struct watch_filter __rcu *filter;
-	struct pipe_inode_info	*pipe;		/* The pipe we're using as a buffer */
+	struct pipe_inode_info	*pipe;		/* Pipe we use as a buffer, NULL if queue closed */
 	struct hlist_head	watches;	/* Contributory watches */
 	struct page		**notes;	/* Preallocated notifications */
 	unsigned long		*notes_bitmap;	/* Allocation bitmap for notes */
@@ -46,7 +46,6 @@ struct watch_queue {
 	spinlock_t		lock;
 	unsigned int		nr_notes;	/* Number of notes */
 	unsigned int		nr_pages;	/* Number of pages in notes[] */
-	bool			defunct;	/* T when queues closed */
 };
 
 /*
diff --git a/kernel/watch_queue.c b/kernel/watch_queue.c
index a6f9bdd956c3..6ead921c15c0 100644
--- a/kernel/watch_queue.c
+++ b/kernel/watch_queue.c
@@ -43,7 +43,7 @@ MODULE_LICENSE("GPL");
 static inline bool lock_wqueue(struct watch_queue *wqueue)
 {
 	spin_lock_bh(&wqueue->lock);
-	if (unlikely(wqueue->defunct)) {
+	if (unlikely(!wqueue->pipe)) {
 		spin_unlock_bh(&wqueue->lock);
 		return false;
 	}
@@ -105,9 +105,6 @@ static bool post_one_notification(struct watch_queue *wqueue,
 	unsigned int head, tail, mask, note, offset, len;
 	bool done = false;
 
-	if (!pipe)
-		return false;
-
 	spin_lock_irq(&pipe->rd_wait.lock);
 
 	mask = pipe->ring_size - 1;
@@ -603,8 +600,11 @@ void watch_queue_clear(struct watch_queue *wqueue)
 	rcu_read_lock();
 	spin_lock_bh(&wqueue->lock);
 
-	/* Prevent new notifications from being stored. */
-	wqueue->defunct = true;
+	/*
+	 * This pipe can be freed by callers like free_pipe_info().
+	 * Removing this reference also prevents new notifications.
+	 */
+	wqueue->pipe = NULL;
 
 	while (!hlist_empty(&wqueue->watches)) {
 		watch = hlist_entry(wqueue->watches.first, struct watch, queue_node);
-- 
2.39.0

Re: [PATCH v4] kernel/watch_queue: NULL the dangling *pipe, and use it for clear check

[David Howells]

Seems reasonable.  Have you run it with the keyutils testsuite?

David

Re: [PATCH v4] kernel/watch_queue: NULL the dangling *pipe, and use it for clear check

[Siddh Raman Pant]

On Wed, 11 Jan 2023 22:07:28 +0530, David Howells wrote:
> Seems reasonable.  Have you run it with the keyutils testsuite?
> 
> David

I had not, but I did now.

I first ran all the keyctl tests with ./runtest.sh keyctl/**/*
All of them passsed. Log: https://pastebin.com/PBm6h5E2

All tests in tests/ pass except features/builtin_trusted, which
fails even without the patch. (Failure log: https://pastebin.com/SGgAbzXp)

Thanks,
Siddh

Re: [PATCH v4] kernel/watch_queue: NULL the dangling *pipe, and use it for clear check

[David Howells]

Siddh Raman Pant <code@siddh.me> wrote:

> All tests in tests/ pass except features/builtin_trusted, which
> fails even without the patch. (Failure log: https://pastebin.com/SGgAbzXp)

Don't worry about that one.  That requires some kernel preparation.

David

Re: [PATCH v4] kernel/watch_queue: NULL the dangling *pipe, and use it for clear check

[Siddh Raman Pant]

On Wed, 11 Jan 2023 23:50:20 +0530, David Howells wrote:
> Siddh Raman Pant code@siddh.me> wrote:
> 
> > All tests in tests/ pass except features/builtin_trusted, which
> > fails even without the patch. (Failure log: https://pastebin.com/SGgAbzXp)
> 
> Don't worry about that one.  That requires some kernel preparation.
> 
> David

Hello,

Please let me know if any changes are required.
Context: https://lore.kernel.org/all/20230111161934.336743-1-code@siddh.me/

Thanks,
Siddh